A0474
Title: Estimation of the number of relevant factors from high-frequency data
Authors: Yuta Koike - University of Tokyo (Japan) [presenting]
Abstract: Factor models play an important role in modeling financial asset prices, both theoretically and practically. Traditionally, only "strong" factors that are correlated with almost all the assets under analysis have been considered, but in recent years, "weak" factors that are correlated with only some assets have attracted attention. It is discussed how to estimate the number of factors that drive the model, including "some" weak factors, from high-frequency data. In particular, a general setting is considered in which the log price process is modeled as a semimartingale, possibly with jumps. Theoretically, the growth rate of the spectral norm of the realized covariance matrix plays a key role, and a new result is given from this perspective.
